Will machine translation be the end of English-only policies in corporations? I believe that machine translation has evolved significantly and already achieved capabilities to replace English-only policies for high-frequency, low-value communications in most languages. For difficult languages, like Chinese, the time is coming soon. The basic reason behind this is that machine translators can now outperform the level of English proficiency that the average person acquires from a typical high school education globally. Thus, it would actually be much cheaper to use translators for almost any kind of business communication.
Machine Translation vs. Incompetent Human Translators
The core of my argument that machine translation has already supplanted the logical need for English-only corporate policies, is found in the performance of machine translators over incompetent translators. It’s no secret today that the majority of translators working for big translation companies (i.e., intermediaries) are incompetent and often produce worse translation quality than that offered by Google Translate — which in itself is already incorrect.
When company employees not native in English, particularly those working overseas, are tasked with producing English versions of documents, their performance is typically close to that of a casual worker translator. That is, a translator who has studied a foreign language and can communicate in it but does not devote enough time to translation to develop useful skills. Thus, if Google Translate can already outperform incompetent translators, it would be reasonable to expect it to also outperform in-house company employees serving as informal, occasional translators. From my own experience reviewing translations provided by a large number of in-house translators, I’ve observed that almost all in-house employees called on to translate are now simply copying and pasting Google Translate, and passing off the translation as their own work.
Using machine translation for international business communications, however, requires special training. In particular, employees need to be well aware that the translation and what is being translated are two separate things entirely. Someone who seems rude on the other end, for example, may not intend to be rude. Furthermore, there is going to be an error rate with machine translation that occasionally renders messages going back and forth unintelligible. Local staff, however, can tell what’s going on because the machine translation will either make no sense, or the conversation itself will make sense but the social context of back-and-forth communication will make no logical sense. In this case, retrying or getting help from actual translators would make more sense. Finally, this approach is better suited to asynchronous communication than conferences.
What if lots of meetings are needed? In this case, I would recommend having certified translators working with interpreters on the content of the presentation. To many companies, the idea of hiring an interpreter to help with corporate meetings and presentations would sound like a terrifying expense. However, consider for a minute that where highly skilled professionals are present, the cost of an interpreter added to the meeting is quite minuscule, as the AIIC points out. Moreover, consider that a top interpreter can enhance an executive’s international leadership efficacy by 10-20% during the meeting. In this case, the value of allowing non-native English-speaking participants fully express themselves and their ideas is far higher than the value of the interpreter. an approach recommended by at least one author in the Harvard Business Review.
Ending English-only policies in favor of formal translation processes can also improve operational efficacy by encouraging employees to place more value on communication. When a company has employees who aren’t professional translators handling such significant volumes of translation work, it creates a belief within the company that the business has low standards for translation and related work. The main reason this occurs is that the use of informal translators is highly indiscriminate: I have seen highly technical reports written in pidgin English by members of leading corporations, that are also totally incorrect. By grading and setting minimum quality standards for professional documents, the corporation tells its international staff that communication is important. Since most serious managerial failures arise from poor communication, this is also a good way to ensure the company’s operational success. Research by Professor Kizito Tekwa of Guangdong Foreign Studies University provides strong evidence that machine translation technology is a highly effective solution for participants, even for Chinese<>English instant message communication.
However, there is a huge caveat on the use of translators and interpreters: they need to be competent and qualified to produce any value. Currently, the majority of translators actually have worse English proficiency than most subsidiaries’ English-speaking staff abroad. Thus, they tend to produce worse results. Many corporations also make the mistake of trying to hire the cheapest translators possible, which only works in fields that are either highly regulated or poor skilled.
Limits of Machine Translation
Nonetheless, machine translation is limited by what kinds of documents it can translate with a reasonable degree of accuracy. Routine text like chat messages and typical e-mails, where participants are aware that machine translation is being used, are places where a machine translator can be expected to outperform amateur human translation. However, the same is not true of professional writings. There is a big difference between professional writing and chats: a chat message usually is written at an elementary school level, is low in complexity, and lacks specialized terminology. On the other hand, professional writings, like contracts, reports, or presentations, generally exceed the capabilities of machine translation programs to accurately handle.
One major reason is the so-called “Garbage In, Garbage Out” principle of artificial intelligence. That is to say, a machine translator is only as good as the quality of work being fed into it, and that generally means the work of incompetent translators. These incompetents have no difficulty handling texts written at a lower level than children’s books, so the machine translators learning from them are likewise able to pick up on that intelligence. But, if it comes to a corporate contract that must be reviewed or a financial analysis of a subsidiary, it would be foolish to rely on a machine translator to handle these kinds of texts. In addition to this “Garbage In, Garbage Out” competency issue, there is another key principle: wise use of staffing resources.
For highly qualified professionals like financial analysts or attorneys, utilizing and making decisions based on translated documents is very expensive. Moreover, the translation will amplify (or destroy) the value of the work that such a person is doing. Assuming they are highly qualified, a translator or interpreter assisting these professionals can significantly enhance the value of their work. This is especially true where lawyers, auditors, and executives are involved. One of these professionals could be costing the company $1,000 an hour, perhaps $10,000 a day even, to resolve an important international matter. Would it make economic sense to throw away all of that value just to save a few hundred dollars on language-related costs? Personally, I have been in meetings that cost $5,000 per hour to run, and none of the English speakers really understood the China-related issue being discussed because employees on the other end of the line in China had no idea how to express such high-level, complex concepts.
Public Policy Dimensions
That covers most of what corporate policy should consider about language, but how should the global advent of machine translation affect public policy for numerous countries? Taking China as an example, which has a language both widely used and very different from English (difficult), I would recommend simply canceling compulsory English education and offering students the option of what language to learn—with computer languages on the menu. If you look at salaries in China by language, computer languages (if included) make more than non-English languages, and English is the least compensated in China. Chinese company Trip.com has actually already begun populating reviews using machine translation from foreign languages and having its customer service staff use machine translation.
In terms of overall economic cost, it would actually be cheaper for a nation to rely on machine translation for chat messages and translators for high-difficulty documents like academic publications. A telling statistic is that China now has about 300,000 translators, but about 2,000,000 English teachers and trainers. In Masters of Translation and Interpreting programs, students have traditionally almost all pursued careers in teaching English because that’s where the jobs are. So, there are about 6 times as many people teaching English as there are actually working as translators. This is particularly surprising when Google Translate could actually translate those materials more accurately! Furthermore, those same people could be learning computer languages, which, unlike English, there is a huge demand for. When you consider that machine translation will be even better in a decade, the idea of universal English education seems like an even bigger waste. China is moving in this direction, as many unemployed English teachers can tell you!
Conclusion
It’s time for machine translation to kill corporate English-only policies and allow employees from different regions to mediate native-language communication using artificial intelligence. The reason this practice has never taken off is, strangely, translation fraud: translators are sending out tons of machine translations they label as human translation. They then caution clients that machine translation is extremely dangerous, and that leads to underutilization of the technology for low-difficulty communications. Using it for professional language production, and not chat messages, is obviously still very dangerous, but machine translation works adequately in conversational contexts where participants are using instant messaging and there is a strong awareness of the potential for the machine to corrupt communications.
